Objective: To establish the prognostic value of knowledge of sentinel node status in melanoma. Design: Single center prospective observational study, with sentinel nodes identified by lymphoscintigraphy, γprobe, and intraoperative blue dye and examined by both conventional histopathology and immunopathology. Setting: Specialist surgical service in west of Scotland. Participants: 482 patients with melanoma who consented to sentinel node biopsy in 1996-2003. Main outcome measure: Time to recurrence of or death from melanoma. Results: Of 472 patients who consented to sentinel node biopsy and in whom at least one sentinel node was identified, 367 (78%) had no tumour in the sentinel node. At mean follow-up of 42 months, 299 (82%) of this group were alive and free from disease, 24 were alive with melanoma recurrence, and 31 had died of melanoma. Of 105 patients with a positive sentinel node biopsy, 44 (42%) were alive and disease free, 12 were alive with recurrence, and 46 had died of melanoma. The survival difference between patients who were negative and those who were positive for tumour in the sentinel node was highly significant at all thickness levels over 1.0 mm (P < 0.001). Multivariate analysis showed that sentinel node status was independent of tumour thickness and ulceration. 71/105 (68%) patients with a positive sentinel node had a negative completion lymphadenectomy, and 44/71 (62%) were alive and disease free at follow-up; 34 patients with a positive sentinel node had further nodes involved, and only 4 (12%) were disease free (P < 0.001).16 patients (13 sentinel node biopsy positive; 3 negative) died of other causes. Conclusion: Sentinel node status is a highly significant predictor of prognosis in melanoma and should be considered in adjuvant studies. However, it should not be regarded as a standard of care until mature data from ongoing randomised trials are available.